ShaniGam

Use encode_features with labels only

2

Is it possible to use encode_features on a model that was trained with `--labels_feat`? I run it and use the labelled images (the same ones I used on training) instead...

Bad results when testing

6

I train the model on 2975 images from the cityscapes dataset with the following command: `python train.py --name label2city_512p_feat_cityscapes --instance_feat` The outputs I get during the training seem fine: ...

warm step

3

Can you please explain the intuition for using `warm_step=200` for only 1 epoch? It doesn't seem like enough for meaningful training without distillation. What happens if I use the distillation...

Great work! I don't really understand how in the unified version you apply the loss on all the detection heads, even when there are no gradients because we didn't pass...

Attempting to use uninitialized value conv2d_1/kernel

9

Whenever I try to start the training I get the error: `FailedPreconditionError (see above for traceback): Attempting to use uninitialized value conv2d_1/kernel`

Bad results when continue training

8

First, thank you for this implementation! I want to continue train a model that I trained before. I run the following command: `python train.py --mixing --init_size 64 ./images/ --ckpt ./checkpoint/140000.model`...